Stage 1 — Provided Ratios Template

What you'll learn

• How a professional financial model separates inputs, assumptions, formulas, and outputs
• What named ranges are and why they make a model portable and specifiable
• Why this project provides the template rather than asking you to build one (hint: it protects time for the specification and AI work)

The architecture

Named ranges — the big idea

A named range assigns a label to a cell, so formulas read like finance instead of like coordinates:

❌ cell-address style:   = B12 / D14
✓ named-range style:    = INC_net_income_2025 / BAL_assets_total_2025

Three consequences, and they carry the whole project:

• Readable — anyone (including an AI) can audit the formula without opening the workbook.
• Portable — formulas survive when the workbook is copied, rows shift, or a new company's data is loaded.
• Specifiable — at Stage 4 you'll write a technical spec that defines every ratio in named-range notation. An LLM that has never seen your Excel file can then compute or verify everything. That's only possible because names, not addresses, are the interface.

Want to feel how the formulas behave? The Ratio Lab mirrors this template's named ranges and averaging conventions — try the sample company.

Two names the valuation stages need

Most CASH_* names are subtotals the ratios use. Two more exist for the DCF in Stages 6 to 8, which needs a base-year free cash flow:

Check that they are there: open your copy → Formulas → Name Manager → both names listed. If they are missing, your copy predates them — download the template again and replace your copy (it is still blank at this stage, so nothing is lost).

The statement tabs show figures in the currency and scale you enter on the Notes tab (Units row) — "VND billion", "EUR million", "USD thousand". Nothing in the template assumes U.S. dollars.

Install the performance-ratios skill

A skill is a folder of instructions an AI agent (Claude Code, Codex, and others) loads on demand. This one is the workbook's named-range contract: what each name means, which valuation inputs can be derived from the names, and the identity checks any model built on the workbook must pass. You run its checks at Stage 3, and the DCF agent in Stages 6 to 8 reads the workbook through it. Install it unmodified at:

[your-repo]/.claude/skills/performance-ratios/SKILL.md

The source is the sample portfolio repo in the course templates folder.

On github.com (no install needed)

1. Open the source link above → click Raw → select all and copy.
2. In your portfolio repo on github.com → Add file → Create new file.
3. In the name box type .claude/skills/performance-ratios/SKILL.md — each / creates a folder.
4. Paste, then Commit changes with the message Add performance-ratios skill.

Or ask Claude Code / Codex

Open it in your repo folder and prompt:

Copy the file at
https://github.com/adamwstauffer/shidler/blob/main/templates/portfolio/firstname-lastname/.claude/skills/performance-ratios/SKILL.md
into .claude/skills/performance-ratios/SKILL.md in this repo, unchanged,
and commit it alone with the message "Add performance-ratios skill".
Show me the commit when you are done.

Read the skill once after installing it. You will not change it at this stage.

What to do

1. Download performance-ratios-template.xlsx from the course templates folder.
2. Open it and click through every tab. Read the Cover tab. Skim the Ratios tab and notice every formula is a sentence in named ranges.
3. Commit an unmodified copy to models/templates/ in your repo.
4. Install the performance-ratios skill, unmodified, at .claude/skills/performance-ratios/SKILL.md.
5. Complete the directory skeleton (if Stage 0 left gaps) — README in every folder.

Set up this project's folders

Stage 0 stood up your portfolio repo. This project adds three folders to that skeleton, and every stage from here commits into one of them. Build them now, each with a one-line README.md inside — Git does not track an empty folder, so the README is what makes the folder exist — and no later stage stalls on housekeeping.

firstname-lastname/
  .claude/
    skills/
      performance-ratios/  SKILL.md, installed unmodified (Stage 1)
  docs/
    decisions/         the selection, spec-iteration and feedback-response memos
  capabilities/
    performance-ratios/  add this: spec.md, the model specification (Stage 4)
  models/
add this whole branch
    templates/         the instructor's template, unmodified
    builds/            your populated workbook
  data/                the filings you worked from, with provenance
  analysis/            the final analysis, the verification, the raw LLM run, the retrospective
  prompt-log.md        your running AI record, at the repo root

Every graded file carries the same prefix — YYYY-MM-DD-{company-slug}, a real date and your company slug, then what the document is (-selection-memo, -spec, -analysis), all lowercase. The repo is named for you; individual files carry the date and company, so a folder full of them stays legible. Your running AI record lives at prompt-log.md at the repo root and is appended to, not replaced. Each stage page restates its own path, so you never have to come back here — this is the whole map in one place.

Submission checklist

• models/templates/performance-ratios-template.xlsx — unmodified, right directory, right filename
• Opened every tab and read the Cover & Instructions tab
• Directory skeleton complete with a meaningful README.md in each directory
• At least 2 new commits since Stage 0 with descriptive messages
• Project folders created with a stub README.md in each
• CASH_capex and CASH_working_capital_change both listed in Name Manager
• .claude/skills/performance-ratios/SKILL.md — unmodified, right directory, right filename

Rubric

The headings below are what the work is read for. Weights and due dates are in your course schedule.

Check yourself